Emerging TechnologyAI & AutomationFalse promises: How marketers can tell the difference between AI and automation

False promises: How marketers can tell the difference between AI and automation

Spotting the difference between AI and automation can be tricky, especially when dealing with tech vendors. Here's an overview of what AI can and can't do.

Artificial intelligence, according to pioneer of the field John McCarthy, is exactly what it sounds like — using a machine to do any task that requires human intelligence. 

So if the saying “So simple, a monkey could do it” applies, then having a machine do it is not AI

However, McCarthy’s definition is also a little too simple. As AI has progressed and technology has become more advanced, explaining exactly how AI works has become more complex as well. 

And because these technologies are so new and difficult to explain, unfortunately there are quite a few impostors out there making promises to marketers that technology can’t keep. 

Here’s how to spot the real from the fake when it comes to choosing AI solutions to solve marketing challenges.  

Content produced in collaboration with Phrasee.

Be wary of companies offering “simple” explanations 

One of the major hurdles many companies face when it comes to helping clients understand their AI solutions is explaining exactly how those solutions work. The design and processes behind AI and machine learning are incredibly complex. 

Plus, the more complex the problem is that an AI has been built to solve, the more complex that AI needs to be. 

AI solutions analyze and process information at speeds and scales that defy human comprehension, so it is also difficult to succinctly explain how and why AI makes the decisions it does, according to Phrasee Chief Scientist Dr. Neil Yager: 

“Simple AI techniques, such as ‘decision trees,’ are nice because they are easy to understand. However, they do not offer the cutting-edge performance more complex techniques do.

On the other hand, more powerful techniques (such as ‘deep learning’) perform very well, but are complex and not easy to understand. There is a trade-off and we’ve decided to optimize our AI towards performance. Anyone who can fully explain their results in simple and intuitive terms is not using state-of-the-art machine learning technology.”

So if an AI salesperson is telling you “It’s simple,” in reference to their solution, those words could be an indicator that the technology is not truly AI. 

What’s the difference between AI and automation?

Many companies sell simple automation as AI, and it can be difficult for a layperson to understand the differences between the two. So businesses might be paying cutting-edge technology prices for solutions that have actually been available for quite some time. 

Like AI, automation uses robotics and rules-based systems to predict outcomes — but those predictions aren’t smart.

For example, automation might allow machine technology to complete repetitive tasks it’s been programmed to complete, but the automation cannot learn from those tasks in order to complete new, related tasks. 

AI, on the other hand, uses the data it analyzes for prediction in order to make new recommendations based on what it’s learned. 

What AI can do

Here’s a list of a few things the best AI technology on the market can do.

  • Understand speech. Smart speakers, like the Amazon Echo, can understand human voices and make recommendations based on commands. 
  • Write compelling copy. If you read the sports section of the newspaper, chances are you’ve read an article written by a robot. AI can quickly analyze relevant keywords and create sentences that approximate the way a human would write. 
  • Give recommendations. Companies like Netflix use AI to analyze massive amounts of data to create personalized recommendations that get even more accurate the more you interact with the platform. 

What AI can’t do

Now here are a few things that AI can’t quite handle yet.

  • Write a script. All those Tweets you may have seen that read something along the lines of “I made a computer watch 1,000 episodes of Law and Order” with hilarious-sounding scripts attached are fake. Actual tests using AI to write scripts have been complete gibberish. 
  • Run an appropriate social media account. When Microsoft used AI to create a Twitter account in the voice of a teenage girl, “Tay,” the account had to be shut down in less than 24 hours. Apparently, what unaided AI “learns” from analyzing thousands of Tweets is how to be racist. Yikes. 
  • Chitchat. Recently, a six-year-old girl got a fun surprise after telling her family’s Alexa Dot about her love of sugar cookies and dollhouses in the form of, you guessed it, nearly $200 worth of sugar cookies and dollhouses. AI can’t yet tell the difference between commands and conversation. 

The evolution of AI has unlocked a world of possibilities, but understanding what is and isn’t possible right now could mean the difference between finding a product that works wonders or falling for something that’s too good to be true. 

To learn more about what’s possible with AI for marketing, check out Phrasee’s report, “Optimizing marketing performance with artificial intelligence.”


US Mobile Streaming Behavior

Whitepaper | Mobile US Mobile Streaming Behavior


US Mobile Streaming Behavior

Streaming has become a staple of US media-viewing habits. Streaming video, however, still comes with a variety of pesky frustrations that viewers are ...

View resource
Winning the Data Game: Digital Analytics Tactics for Media Groups

Whitepaper | Actionable Analysis Winning the Data Game: Digital Analytics Tactics for Media Groups


Winning the Data Game: Digital Analytics Tactics f...

Data is the lifeblood of so many companies today. You need more of it, all of which at higher quality, and all the meanwhile being compliant with data...

View resource
Learning to win the talent war: how digital marketing can develop its people

Whitepaper | Digital Marketing Learning to win the talent war: how digital marketing can develop its people


Learning to win the talent war: how digital market...

This report documents the findings of a Fireside chat held by ClickZ in the first quarter of 2022. It provides expert insight on how companies can ret...

View resource
Using storytelling for strategic planning

Podcast | Strategy Using storytelling for strategic planning


Using storytelling for strategic planning

Each story requires not only a hero but an adversary to overcome and a beneficiary who will share the goal of the hero. Whilst these characters can be...

View resource